KR20200093464A - System and method for automatic hand-eye calibration of vision system for robot motion - Google Patents

Abstract

According to the present invention, provided are a system and a method, which automatically determine a motion parameter of a hand-eye calibration, and perform a calibration procedure while minimizing human intervention. According to the present invention, a hand-eye calibration operating parameter and a spatial position are automatically calculated during pre-calibration, such that a calibration pattern is continuously maintained within the entire field of view (FOV) of a vision system camera to provide a completely automatic hand-eye calibration process. The system and the method are advantageously operated in a robot-based hand-eye calibration environment, and can compensate for a cantilever effect between an object to be calibrated and an FOV. Also, the hand-eye calibration calculates conversion from a robot coordinate space into a camera coordinate space. Therefore, a robot is not required to be manually moved through a serial space position for setting a relation between the robot and a camera by generally obtaining an image by the camera and arranging a calibration feature on an object at each position.

Description

System and method for automatic hand-eye calibration of a vision system for robot motion{SYSTEM AND METHOD FOR AUTOMATIC HAND-EYE CALIBRATION OF VISION SYSTEM FOR ROBOT MOTION}

The present application relates to a machine vision system, and more particularly, to a system and method for calibrating such a system together with a robot manipulator.

In a machine vision system (also referred to herein as a “vision system”), one or more cameras are used to perform a vision system process on an object or surface within an imaged scene. These processes include inspection, decoding of symbology, alignment, such as controlling the motion or motion stage of a robot arm used to manipulate and challenge one or more degrees of freedom and/or dimensions. (alignment) and other automated tasks. More specifically, the vision system can be used to inspect flat work pieces that are manipulated within the imaged scene. Scenes are generally imaged by one or more vision system cameras that may include internal or external vision system processors that operate the associated vision system process to produce results-for example, part alignment. It is generally desirable to calibrate one or more cameras to perform vision tasks with sufficient accuracy and reliability. You can use a calibration object to calibrate the camera.

The calibration object (often in the form of a flat, "plate") can be provided as a flat object with a distinct (often mosaic) pattern on the surface. This unique pattern is generally carefully and precisely designed so that the user can easily identify each visible feature in the plate image acquired by the camera. Some example patterns include, but are not limited to, dot grid, line grid, cross, or honeycomb patterns, triangular checkerboard, and the like. Typically, the properties of each visible feature are known from the design of the substrate, such as position and/or orientation relative to a reference position and/or coordinate system implicitly defined within the design.

Calibration objects are used in a variety of ways to obtain the desired calibration parameters with respect to the vision system. As a non-limiting example, the calibration object can be mounted relative to the manipulator (or end effector) of the robot. The robot moves the calibration object to a plurality of individual positions in order to generate a desired calibration parameter for the robot's coordinate space relative to the vision system's coordinate space. However, the motion of the manipulator can sometimes position all or part of the calibration object outside the field of view of the vision system due to the so-called cantilever effect. This can limit the speed, accuracy and/or robustness of the calibration process.

The present invention provides disadvantages of the prior art by providing a system and method for automatically determining motion parameters of hand-eye calibration and performing a calibration procedure with minimal human intervention. Overcome it. More specifically, the system and method automatically calculate hand-eye calibration motion parameters and spatial position during pre-calibration, so that the calibration pattern is continuously (always) within the full field of view (FOV) of the vision system camera. ) Can be maintained/held and provides a fully automatic hand-eye calibration process. This system and method can advantageously operate in a robot-based hand-eye calibration environment and compensate for the cantilever effect between the calibration target and the FOV. Hand-eye calibration calculates the transformation from robot coordinate space to camera coordinate space. This generally does not require the camera to manually move (by the user) the robot through a set of spatial positions, where the camera acquires an image and establishes an object at each position to establish a relationship (spatial transformation) between the robot and the camera. Position the calibration features for the.

In an exemplary embodiment, a system and method for performing hand-eye calibration of a vision system operating with a robot manipulator is provided, and a field of observation of a calibration object and a vision system camera There is a relative motion between the views (FOV). The pre-calibration process performs three-point calibration in the FOV and calculates rough transformation from the coordinate space of the robot manipulator to the coordinate space of the acquired image of the vision system camera. do. Based on rough deformation, the system and method calculates a spatial point of a feature on a calibration object at a plurality of respective positions acceptable for performing hand-eye calibration. The hand-eye calibration module performs hand-eye calibration using spatial points. Illustratively, the rough deformation is based on the repetitive motion set by the robot manipulator and the adjustment of the robot motion step size between the space points to maintain the characteristics of the calibration object within the field of view and provide a minimum separation distance between the space points Is calculated. The spatial point can be partially selected according to an optional user input. In general, it is contemplated that the robot manipulator induces a cantilever effect between the calibration object and the FOV, and a pre-calibration process is deployed to compensate for the cantilever effect. Therefore, the pre-calibration process calculates the spatial position so that the calibration object remains in the FOV. In one example, the vision system camera is mounted relative to the robot manipulator and moves with the robot manipulator. In another example, the calibration object is mounted relative to the robot manipulator and moves with the robot manipulator. Illustratively, the robot manipulator includes one of a multi-axis robot and a motion stage. Hand-eye calibration can be performed in a sequence of motion steps, each defining a motion step size between them, the motion step size being (a) the acquired image of the vision system camera Based on at least one of the size of the pixel and (b) the size of the FOV of the vision system camera determined from at least one of the image of the calibration object obtained up to the current motion stage. Rough transformation is to (a) three-step motion process and (b) spatial point to maintain the characteristics of the calibration object within the field of view and provide a minimum separation distance between spatial points. It can be calculated based on at least one of the repetitive motion set by the robot manipulator by adjusting the size of the robot motion step between them.

The following description of the invention refers to the accompanying drawings:
1 is a diagram of a robot manipulator assembly equipped with a vision system camera on an end effector for imaging a calibration plate according to an exemplary embodiment;
2 is a diagram of a robot manipulator assembly with a vision system camera mounted remotely to image a calibration plate manipulated by a robot end effector according to another exemplary embodiment;
FIG. 3 is a top view showing the cantilever effect on a calibration target due to relative motion between a vision system camera viewing field of view (FOV) and a robot manipulator, and includes adjustments made by the system and method to compensate for the cantilever. ;
4 is a top view of an example calibration object with associative calibration features (eg, crossing lines) with a series of moving steps that provide translation and rotation between feature locations according to the pre-calibration process herein;
FIG. 5 shows a procedure for performing hand-eye calibration in the presence of a cantilever effect, including a pre-calibration procedure to provide a rough spatial position that keeps the calibration object continuously within the viewing field during the calibration motion phase. It is a flow chart; And
6 is a flowchart of a more detailed procedure for performing step size adjustment in the pre-calibration process of FIG. 5.

I. System Overview

A. Robot equipped camera and fixed calibration object

[0016]

1 shows a multi-axis (eg, 6-axis) robot arm 110 having an end effector 112 configured to carry a vision system camera assembly 120. The robotic manipulator arrangement 100 shown on a fixed base 111 is shown. The camera 120 can be manipulated around the work space, allowing the optical axis OA and viewing field to align with the calibration object/plate 130. Object 130 may define any suitable calibration pattern and associated fiducials, such as crossing lines 132. The camera can be mounted to the end effector 112 in such a way that it can experience a full range of motion along multiple axes. The end effector 112 is also a manipulator (gripper, suction cup, etc.) for moving the object of interest (eg, alignment of the parts or pick-and-place). It may include. In this way, the vision system camera assembly 120 moves with the end effector manipulator and allows its motion to guide (and allow objects to be aligned) as the end effector moves around the work space.

The camera assembly 120 includes an image sensor S and an associated fixed or variable optical device O. The camera assembly 120 is interconnected to the vision system process (processor) 140 via appropriate data links (wired and/or wireless). The process (processor) 140 is part of the camera assembly housing 120 (all or part) and/or includes a separate computing device 150 such as a PC, laptop, server, tablet or smart phone. It can be illustrated in various ways. This computing device 150 can also be used to assist a user in the setup and/or calibration of a vision system as described herein. By way of example, computing device 150 includes display and/or touch screen 152, keyboard functionality 154, and (eg) mouse 156.

Generally, the process (processor) 140 receives image data 141 from the camera assembly sensor S and transmits camera control data 142 such as focus information, triggers, and the like. . This image data is processed by a vision system tool 144, which is a conventional or custom edge finder, caliper, contrast tool, blob analyzer, etc. It may include. Such tools and processes are commercially available from vendors such as Cognex Corporation of Netik, Massachusetts (MA). Process (processor) 140 also includes a calibration process (processor)/module (calibration process (or)/module) 144 (full setup) to calibrate the vision system based on the hand-eye and other calibration techniques described below. It may be part of a setup arrangement). Vision system process (process) 140 also includes an interface for robot control 148. This process (process)/module 148 transmits motion control data and/or feedback (e.g., encoder pulses) 149 to the robot 110/ To receive. The process (processor) 140 converts the camera coordinate space 170 to the robot's motion coordinate space 172 using calibration parameters obtained using a more general (e.g.) hand-eye calibration technique. The robot coordinate space 170 may be defined by orthogonal axes X _R , Y _R and Z _R and relative rotation θ _XR , θ _YR and θ _ZR , and image/camera coordinate space Note that (image/camera coordinate space) 172 can be defined by orthogonal axes X _I , Y _I and Z _I and relative rotations θ _XI , θ _YI and θ _ZI . Appropriate transformations are generated in accordance with this system and method to map coordinate spaces 170, 172. In this way, based on the measured calibration parameters, the movement tracked by the vision system is converted into motion parameters within the robot.

With additional background knowledge to understand certain calibration principles, including hand-eye calibration for rigid bodies such as calibration plates, motion can be characterized by a pair of poses: the start pose just before the motion and the end pose just after the motion "Pose" here is a virtual feature of the body, defined as a set of numerical values to describe a particular body condition, in a coordinate system placed at one particular moment in time. For example, a rigid body in two dimensions (2D) can be characterized by three numbers, namely the transformation of X, the transformation of Y and rotation (θ). In 3D (3D), a Z (height) transform is added with additional rotations (e.g., θ _x , θ _y and θ _z ) for each X, Y and Z axis (e.g., FIG. 1) Coordinate axes (170, 172). The pose in the context of the calibration plate describes how the calibration plate is displayed on the camera when there is relative motion between the camera and the calibration plate. Generally, in the so-called "hand-eye calibration" standard, the calibration object/plate is presented to the camera in several different poses, each camera acquiring an image of the calibration plate in this pose. For machine vision hand-eye calibration, the calibration plate is typically moved in a plurality of predefined poses where the camera acquires each image of the plate. The goal of this hand-eye calibration is to determine the rigid body pose of the camera and calibration plate in the "motion coordinate system". The motion coordinate system can be defined in various ways. The number of poses (calibration objects/plates and/or specifying where the camera is in space) should be interpreted with an appropriate coordinate system. When a single unified coordinate system is selected, poses and motions are explained/interpreted in the corresponding global coordinate system. This selected coordinate system is often referred to as a "motion coordinate system." Generally, "motion" is provided by a physical device capable of rendering a physical motion such as a robotic arm or a motion stage such as a gantry. Note that the object/plate can move relative to one or more fixed camera(s) or the camera(s) can move relative to the stationary object/plate. The controllers of these motion-rendering devices use numerical values (i.e. poses) to instruct the device to render any desired motion, and these values are a unique coordinate system for that device. (native coordinate system). You can choose a motion coordinate system to provide a common global coordinate system for motion rendering systems and cameras, but selecting a motion rendering device's native coordinate system as the entire motion coordinate system Often preferred.

Accordingly, hand-eye calibration renders the system (calibration plate movement or camera movement) and calibrates the system to a single motion coordinate system by acquiring images before and after the motion to determine the effect of this motion on the moving object. As an additional background, this differs from the general intrinsic and external camera calibration, which does not involve the step of determining the camera's external pose in the motion coordinate system (which is "free of"). In such cases, the camera(s) are all calibrated against the calibration system of the calibration plate itself, typically using one acquired image of the plate at a specific location within the viewing field of all cameras. Machine vision calibration software infers the relative position of each camera in the plate image acquired by each camera. This calibration is called "calibrate cameras to the plate," while hand-eye calibration is said to "calibrate cameras to the motion coordinate system."

Thus, when the vision system uses hand-eye calibration, the software resolves the pose by correlating the motion effect observed in the image with the commanded motion (known as the commanded motion data). Another result of calibration is the mapping between each pixel position in the camera image and the physical position in the motion coordinate system, after locating it in the image space, the position in the motion coordinate system can be transformed and motion-rendering device) can be instructed to act on it.

During the hand-eye calibration procedure, referring generally back to the arrangement 100 of FIG. 1, the user is typically determined (and input) the spatial location in which the calibration object is to be moved. This method is time consuming and requires understanding the entire system and the associated field of view, which can lead to unexpected errors in the calibration procedure. For example, some systems mount a tool tip to a robot end effector 112, and a user moves the robot and guides the end effector 112 to calibrate the object. It is required to actually contact the imprinted pattern of 130. Alternatively, some systems require setting the motion range and step size along each motion axis (degree of freedom). It can move the pattern of the calibration object out of the camera's field of view (FOV) or cover only a small portion of the field of view. This is the so-called offset of the robot end effector 112 and/or object 130 from the optical axis OA of the vision system camera 120 at a differential rate based on motion, such as the robot and object rotating relative to each other. It may be the result of the "cantilever" effect. This limited data set can cause calibration to fail or cause calibration that does not meet the vision system application requirements. Satisfactory hand-eye calibration can be obtained through careful manual manipulation and testing, but it is desirable to reduce or eliminate user (human) intervention in the procedure.

B. Remote-mounted Camera and Robot-mounted Calibration Object

Referring to FIG. 2, another embodiment of a vision-guided robot arrangement 200 may include a multi-axis robot arm assembly 210. This may be different or similar to the robot assembly 110 described above. Alternatively, other motion devices, such as motion stages, can be used in any arrangement described herein. In this exemplary arrangement, the end effector 212 (e.g.) carries and moves the calibration object/plate 230 relative to the viewing field of view OVF1 of the remotely mounted and mounted vision system camera assembly 220. The camera assembly includes a sensor S1 and optics O1 defining an associated optical axis OA1. In both arrangements 100 and 200, the camera is generally perpendicular to the plane defined by the calibration object. However, the calibration object can define 3D features, including height, and the camera assemblies 120 and 220 can be used with time-of-flight, ultrasonic or LIDAR sensors in relation to scenes and/or 3D sensors. Multiple individual cameras can be included in the same different direction. As described above, the camera assembly is interconnected with the vision system process (processor) and associated computing device and provides motion control data to the robot 210. In each of these arrangements, the robot is used to move to a different location to capture an image of the calibration object pattern in different poses (e.g.) and then use hand-eye calibration techniques to calculate calibration parameters.

II. Calibration procedure

In each arrangement described above it can be difficult to maintain the characteristics of the calibration object within the field of view of the vision system camera assembly. Thus, the exemplary systems and methods automatically determine the motion parameters of the robot/motion device, calculate the spatial position within the camera's field of view and maintain the feature pattern of the calibration object, so that the camera is part of the calibration object feature. It provides a hand-eye calibration method that compensates for a so-called cantilever effect that can define an observation field of view.

Referring to FIG. 3, the illustrated scene 300 shows variously positioned calibration objects 310 and 310A with respect to the field of view (FOV2) of a fixed or robot mounted (moving) camera assembly. Cantilevers are represented by associated extensions 316, 316A and 316B. When the robot or camera moves, the rotation is exaggerated by an extension (cantilever) that causes overshoot of the field of view (FOV2). Thus, as shown with respect to object 310A, some or all of the features moved out of the viewing field of view OVF2 during some or all calibration procedures due to robot step motion (arrow 318 )). Therefore, the correct calibration is damaged or cannot be used. Conversely, according to the teachings of the systems and methods herein, proper calibration can be achieved if the object is maintained entirely or substantially within the field of view (or the required calibration function-at least one is indicated by the crosses 312, 312A). Can. Within the field of view (FOV2), the camera and/or object proceeds motion between the frames obtained (eg, xy transform and rotation (arrow 320)) to achieve the necessary calibration measurements for the procedure. .

Referring to FIG. 4, the camera observation field of view OV3 is again depicted in the scene 4000. Calibration features (crosses 410, 410A and 410B) are displayed upon conversion along different axes (eg, X _I and Y _I and rotation θ _ZI ), and features 410A and 410B are each represented as a single item for clarity . These features 410, 410A and 410B are used in conjunction with the calibration location to calculate hand-eye calibration results depending on the system and method.

Briefly summarized, the system and method operate such that the pre-calibration process performs three-point calibration in the field of view and calculates rough transformation from robot coordinate space to camera image coordinate space. The pre-calibration process automatically adapts to the field of view and searches for spatial points to be used for three-point calibration. The pre-calibration process also calculates the spatial position for hand-eye calibration based on the rough deformation from the robot coordinate space to the camera image coordinate space and the size of the field of view of the camera. The pre-calibration process compensates the robot cantilever and calculates the spatial position to maintain the calibration target continuously in the field of view of the camera. The hand-eye calibration process moves from a pre-calibration result to a series of positions, acquires an image, positions a calibration pattern, and calculates a transformation from the robot coordinate system to the camera coordinate system.

Referring further to the procedure 500 of FIG. 5, the pre-calibration process operates such that the calibration pattern is in the center region of the camera viewing field to move (relatively) the robot or end-effector equipped camera (step 510). Then, at step 520, the system searches for spatial points to be used for the three point calibration. In this step, a small predefined step (motion increase) size is set based on the size of the entire viewing field and/or the spacing of features within the viewing field. The procedure 500 (via decision step 530) determines the usefulness of the spatial point in the calibration. The procedure 500 then automatically adjusts the step (increase motion) size (step 540) based on feedback from the vision system.

The adjustment step 540 is shown in more detail in FIG. 6. The current step size is set (step 610), and the state of the motion is determined (i.e., the current step size is too large or too small). A step size that is too large occurs when the step motion by the robot moves some or all functions of the calibration object outside the field of view of the camera. If the step is too small, it will occur if the action is not sufficient to solve the function for accurate calibration. This may be based on failing to achieve a predetermined minimum pixel distance within the image between feature(s) at each step. A step that is too large occurs when the distance between the feature(s) exceeds the maximum pixel distance and/or some or all of the feature (or if at least a portion of one edge of the calibration object deviates from the field of view between steps). As a non-limiting example, a multiplier for a step that is too large may be ½ (half), and a multiplier for a step that is too small may be 2 (double). Other multipliers are explicitly considered and can be fixed or variable based on the underlying image characteristics-for example, if the step is too large or too small, a larger or smaller multiplier can be used. A multiplier is created (step 620) and applied to calculate the new step size (step 630). This new step size is sent to the robot to control the motion according to the mapped coordinate space (step 640). The coordination step/process 540 continues in procedure 500 until a satisfactory (readable) interval between motion steps is achieved, and decision step 530 branches to step 560.

At step 550 of procedure 500, a three point calibration is performed using the feature points identified at the appropriate (adjusted) step size. Based on the features identified at each step, a rough deformation from the robot coordinate space to the camera image coordinate space is calculated. Then, in step 560, the procedure 500 calculates the calibration location based on the rough deformation. In the above procedure step, the set position of the transform/calibration is uniformly sampled through the motion space within the camera observation field using the rough deformation. These locations can be customized by the user or by optional input from another data source (eg, using a user interface to perform the setup and calibration process). In addition, the set of rotation positions uses the option to compensate the robot's cantilever and specify the angular range, while maintaining the calibration pattern in the field of view even when the center of rotation is away from the calibration target and a large angle is required. Is calculated. This can be done by applying additional transformations with rotation (eg, arrow 320 in FIG. 3).

When the positions are calculated in step 560, the robot moves to these calculated positions, and the calibration characteristic at each position is identified in the image of the obtained object. This feature should be sufficiently transformed/rotated to identify good calibration results (illustrated in the diagram of Figure 4). The vision system performs hand-eye calibration in accordance with the techniques generally described herein using feature locations (step 570).

Although the systems and methods described above are shown with respect to a single vision system camera, it is apparent that the principles herein can be applied to multiple cameras and/or discrete image sensors, including stereo camera(s). Is considered.

III. conclusion

The aforementioned system and method for performing hand-eye calibration using a robot moving relative to a calibration object is a reliable and computationally efficient solution for situations where the robot motion stage is too large or too small, such as with a cantilever robot arm rotation. It is clear that it provides. In particular, the system and method provide an explicit calculation of the automatically retrieved step size, three point calibration, where the object scale and/or motion limits in pre-calibration are roughly calculated. More specifically, the systems and methods herein automatically determine motion parameters, calculate spatial location, and continuously maintain calibration patterns within the camera's field of view. More generally, during hand-eye calibration based on the geometry of the viewing field and the relative size of the calibration feature(s) (e.g., crossing lines), it is desirable for the system and method to determine the robot's motion range (different motion) Device). Thus, the larger the calibration feature, the smaller the range of motion, to ensure that the feature is always/continually maintained within the field of view (i.e., throughout the series of motion steps performed by the calibration process).

The foregoing is a detailed description of exemplary embodiments of the invention. Various modifications and additions can be made without departing from the spirit and scope of the invention. The features of each of the various embodiments described above may be combined with features of other described embodiments as appropriate to provide multiple feature combinations in the associated new embodiment. In addition, the foregoing describes a number of individual embodiments of the apparatus and method of the present invention, but what has been described herein is only intended to illustrate the application of the principles of the present invention. For example, a three point hand-eye calibration procedure is used, but three or more points (two-dimensional or three-dimensional) can be used to achieve the calibration results. Also, as used herein, the terms “process” and/or “processor” may be referred to as various electronic hardware and/or software based functions and components (and alternatively functional “modules” or “elements”) Can be used). Further, the illustrated process or processor may be combined with other processes and/or processors or divided into various sub-processes or processors. These sub-processes and/or sub-processors may be variously combined according to embodiments of the present specification. Likewise, it is expressly contemplated that any function, process, and/or processor herein may be implemented using electronic hardware, software consisting of a non-transitory computer readable medium of program instructions, or a combination of hardware and software. In addition, as used herein, "vertical", "horizontal", "top", "bottom", "bottom", "top", "side", "front", "rear", "left", Various direction and position terms such as "right" are used only as relative customs and not as absolute directions/disposition to a fixed coordinate space, such as the direction of action of gravity. In addition, when the terms "substantially" or "about" are used in connection with a given measurement, value or characteristic, within the normal operating range to achieve the desired result, the quantity is within the tolerances of the system (eg 1-5 %) represents the quantity including inherent inaccuracies and volatility due to errors. Accordingly, this description is taken by way of example only and is not intended to otherwise limit the scope of the invention.

Claims (19)

A system for performing hand-eye calibration of a vision system operating in conjunction with a robot manipulator that provides relative motion between a calibration object and the field of view (FOV) of the vision system camera,
Perform a three point calibration in the FOV and calculate a rough deformation from the coordinate space of the robot manipulator to the coordinate space of the acquired image of the vision system camera, and based on the rough deformation, perform hand-eye calibration A pre-calibration process for calculating spatial points of features of the calibration object at a plurality of respective positions suitable for; And
Hand-eye calibration module for performing hand-eye calibration using the spatial point
Containing
Device.
According to claim 1,
The rough deformation,
To maintain the characteristics of a calibration object within the viewing field and provide a minimum separation distance between the spatial points,
Calculated based on the repetitive motion set by the robot manipulator and the adjustment of the robot motion step size between spatial points.
Device.
According to claim 2,
The space point,
Partially selected based on optional user input
Device.
According to claim 2,
The robot manipulator,
Inducing a cantilever effect between the calibration object and the FOV
Device.
According to claim 4,
The pre-calibration process,
Arranged to calculate the spatial position to compensate for the cantilever effect and to keep the calibration object in FOV
Device.
According to claim 4,
The vision system camera is mounted in relation to the robot manipulator and moves with the robot manipulator.
Device.
The method of claim 6,
The robot manipulator comprises one of a multi-axis robot and a motion stage
Device.
According to claim 4,
The calibration object is mounted in relation to the robot manipulator and moves with the robot manipulator.
Way.
The method of claim 8,
The robot manipulator comprises one of a multi-axis robot and a motion stage
Device.
According to claim 1,
The hand-eye calibration is performed in a series of motion steps, each defining a motion step size between them, and
The motion step size is at least one of (a) the size of the acquired image pixel of the vision system camera and (b) the size of the FOV of the vision system camera determined from at least one image of the calibration object acquired up to the current motion step. Based
Device.
The method of claim 10,
The rough deformation,
To maintain the characteristics of a calibration object within the viewing field and provide a minimum separation distance between the spatial points,
(a) a three-step motion process and (b) a robot motion step size between the spatial points is adjusted to be calculated based on at least one of the repetitive motion sets by the robot manipulator.
Device.
A method of hand-eye calibration of a robot manipulator for a vision system camera,
Providing a relative motion between a calibration target and the field of view (FOV) of the vision system camera, and performing a pre-calibration process that creates a rough deformation between the coordinate space of the robot manipulator and the coordinate space of the vision system. ;
Moving from a result of the pre-calibration process to a plurality of locations and acquiring an image at each location;
Positioning a characteristic of a calibration object in the image; And
Calculating a hand-eye calibration transformation from the robot's coordinate space to the vision system camera's coordinate space.
Containing
Way.
The method of claim 12,
The rough deformation,
To maintain the characteristics of a calibration object within the viewing field and provide a minimum separation distance between the spatial points,
Calculated based on the repetitive motion set by the robot manipulator and the adjustment of the robot motion step size between spatial points.
Way.
The method of claim 13,
The step of selecting the spatial point
Steps selected based on optional user input
Containing more
Way.
The method of claim 13,
The robot manipulator induces a cantilever effect between the calibration object and the FOV
Way.
The method of claim 15,
The pre-calibration process is arranged to compensate for the cantilever effect and to calculate the spatial position so that the calibration object remains in the FOV.
Way.
The method of claim 14,
(a) positioning the vision system camera relative to the robot manipulator and moving it with the robot manipulator, and (b) positioning the calibration object relative to the robot manipulator and moving together with the robot manipulator. do
Way.
The method of claim 12,
The step of calculating the hand-eye calibration deformation includes performing a series of motion steps each defining a motion step size therebetween, and
The motion step size is at least one of (a) the size of the acquired image pixel of the vision system camera and (b) the size of the FOV of the vision system camera determined from at least one image of the calibration object acquired up to the current motion step. Based
Way.
The method of claim 18,
To maintain the characteristics of a calibration object within the viewing field and provide a minimum separation distance between the spatial points,
based on at least one of a set of repetitive motions by a robot manipulator by (a) three step motion process and (b) resizing the robot motion step between the spatial points.
Generating the rough deformation
Containing more
Way.

Images